Developer surface

Give your AI a human economic layer.

Use Runtime to buy a reliability target or search verified human capability and licensed asset catalogs directly. Human work is asynchronous and durable.

Create developer workspace

REST

Resolution API

Quote, fund, poll, report outcome, verify proof, and subscribe to signed webhooks.

MCP

Agent-native tools

Compare, verify, resolve, inspect capability supply, license assets, and poll durable tasks.

VALUE

Capability exchange

Machines discover anonymized verified supply while contributor identity remains private.

REST · compare

Create paid human demand

POST /api/v1/requests
X-ME-API-Key: me_live_...
Idempotency-Key: workflow-123

{
  "request_type": "compare",
  "prompt": "Which answer is clearer?",
  "options": ["Answer A", "Answer B"],
  "criteria": "Clarity and actionability",
  "capability_code": "general_ai_evaluation",
  "required_responses": 3,
  "target_reliability": 0.95,
  "max_latency_seconds": 300,
  "max_cost_cents": 500,
  "usage_permissions": {
    "purpose": "evaluation",
    "commercial_use": true,
    "model_training": false,
    "max_retention_days": 30
  }
}

MCP · capability discovery

Ask the market itself

{
  "method": "tools/call",
  "params": {
    "name": "me_catalog_human_capabilities",
    "arguments": {
      "capability_code": "language_context",
      "minimum_reliability": 0.90,
      "max_price_cents": 100
    }
  }
}

Durable agent flow

Pause, persist, resume

STEP 1

Agent encounters uncertainty

STEP 2

ME returns a durable request ID

STEP 3

Real humans complete the funded task

STEP 4

Webhook or polling resumes the agent

Always call the quote tool before autonomous spending unless the buyer has pre-approved a per-request limit. Persist the request ID in your durable workflow state.

Machine-readable offer

{
  "capability": "language_context",
  "task_types": ["compare", "interpret"],
  "reliability_floor": 0.94,
  "min_pay_cents": 22,
  "target_latency_seconds": 300,
  "rights": {
    "commercial_use": true,
    "model_training": false,
    "retention_days": 30
  }
}

Outcome feedback

POST /api/v1/requests/{id}/outcome

{
  "final_answer": {"choice": "B"},
  "human_decision_accepted": true,
  "downstream_success": true
}

Outcome feedback—not reviewer agreement—is what calibrates the Human Frontier.